Method for diagnosing skin condition using artificial intelligence and skin beauty device using the same

Abstract

The present invention relates to a skin beauty device using artificial intelligence, comprising: an image receiving module that receives information on a user's face image and information on shooting environment conditions from a user terminal; a sensor module that collects at least one of an optical signal, a temperature signal, and an impedance signal from a skin surface; an on-device artificial intelligence module that is electrically connected to the image receiving module and the sensor module, calculates a skin condition feature using the face image information and the signal collected by the sensor module, and infers a stimulation parameter corresponding to the calculated skin condition feature; a stimulation module that is electrically connected to the on-device artificial intelligence module and outputs at least one of electrical stimulation, photo stimulation, high-frequency stimulation, and thermal stimulation according to the stimulation parameter; and a standardized metadata storage module that is electrically connected to the image receiving module, the sensor module, and the on-device artificial intelligence module, and, in order to compare the effect of improving the user's skin condition, generates standardized metadata by standardizing each of the face image information, the shooting environment condition information, the optical signal, the temperature signal, and the impedance signal according to an analysis reference value, and stores the generated standardized metadata. and includes a cosmetic recommendation module that is electrically connected to the on-device artificial intelligence module and generates a list of cosmetics suitable for the user's skin according to the skin condition feature and provides it to the user terminal, wherein the on-device artificial intelligence module sets individual skin condition reference features using the face capture image information and adjusts the stimulation parameter in a real-time closed-loop manner by comparing the skin condition feature after stimulation is applied to the skin surface by the stimulation module with the individual skin condition reference feature, and estimates the skin response pattern using a curve fitting algorithm, and the stimulation intensity of the stimulation parameter according to the estimated skin response pattern,The invention is characterized by automatically optimizing the stimulation time and stimulation waveform, predicting the possibility of overheating, overstimulation, and pain based on the impedance signal and the temperature signal to adjust the stimulation parameters, and automatically stopping the stimulation output or lowering it to within a safe limit when contact failure or operational deviation is detected, and the curve fitting algorithm adjusts the stimulation parameters by approximating the skin response to skin stimulation with a curve, and performs closed-loop learning to adapt to changes in individual skin sensitivity by iteratively correcting the curve using real-time skin response, and the cosmetic recommendation module receives information regarding the user's total income and fixed expenses from the user terminal to calculate disposable income, first filters cosmetics suitable for the user's skin according to the skin condition features, and secondarily filters cosmetics among the first-filtered cosmetics that are less than or equal to a preset ratio of the disposable income to generate the cosmetic list, wherein the preset ratio is adjusted according to the user's gender and age group. According to the present invention, the on-device artificial intelligence module learns individual skin condition standard features and enables real-time setting of optimal stimulation according to the user's skin characteristics and session environment. Since stimulation parameters are adjusted in a closed-loop manner, it is effective in applying skin stimulation optimized for the user.

Description

Technology Field

[0001] The present invention relates to a skin condition diagnosis method using artificial intelligence and a skin beauty device using the same. More specifically, it relates to a skin condition diagnosis method and a skin beauty device using the same that can apply optimal stimulation according to the user's skin condition by analyzing the user's skin surface condition using artificial intelligence and inferring stimulation parameters suitable for the user's skin in a real-time closed-loop manner based on the analysis results. Background Technology

[0002] With the recent spread of trends seeking both health and aesthetic satisfaction, interest in skin care is rapidly increasing. Consequently, various types of portable skin care devices are being developed that can be easily used by users at home, as well as in dermatology clinics and specialized skin clinics.

[0003] However, most existing skin care devices operate using fixed presets or simple timers, which limits their ability to reflect changes in the user's skin condition or environmental conditions in real time.

[0004] For example, methods of controlling the intensity or duration of electrical stimulation (EMS), light stimulation (LED), and radio frequency stimulation (RF) to preset values ​​have been simple and widely used, but the stimulation may be excessive or insufficient depending on the user's skin moisture content, blood flow, temperature, impedance, or the state of the skin's color and texture. Such fixed control fails to provide optimal stimulation according to skin condition, resulting in large variations in effectiveness and insufficient safety regarding skin irritation.

[0005] In addition, most existing skin beauty devices rely on static analysis based on a single image or are implemented with a structure that performs computations on a cloud server, which results in slow analysis speeds depending on the network environment or poses a risk of personal information leakage. Furthermore, there is a problem in that the computational load and memory requirements are excessive, making them unsuitable for installation in portable or home beauty devices.

[0006] Meanwhile, skin condition is complexly determined by various factors such as brightness, erythema, yellowing, wrinkles, pores, impedance, and temperature, and the shooting environment, including lighting intensity, ISO sensitivity, white balance, and focal length, also significantly affects measurement results. However, existing skin beauty devices lack sufficient correction procedures for these external variables, making it difficult to quantitatively compare skin conditions between sessions and reliably evaluate skin improvement effects. Consequently, even for the same user, measurement data varies significantly with slight changes in the shooting environment, which lowers the learning accuracy of artificial intelligence models.

[0007] Accordingly, there is a growing need for an on-device AI-based skin beauty device that simultaneously ensures rapid responsiveness and data security through a lightweight AI architecture that eliminates cloud dependency, enables long-term tracking of skin improvement trends using standardized session metadata, calculates optimal stimulation conditions in real time by quantifying skin condition through comprehensive analysis of facial images and biosignals, and allows for the immediate adjustment of stimulation parameters based on skin condition analysis via a real-time closed-loop control method by learning stimulation-response relationships based on long-term user data. Prior art literature

[0008] Republic of Korea Published Patent No. 10-2024-0044555 The problem to be solved

[0009] The objective of the present invention is to solve the aforementioned conventional problems by providing a skin condition diagnosis method and a skin beauty device using the same, which can apply optimal stimulation according to the user's skin condition by analyzing the user's skin surface condition using artificial intelligence and inferring stimulation parameters suitable for the user's skin in a real-time closed-loop manner based on the analysis results. means of solving the problem

[0010] The above objective is achieved, according to the present invention, by an image receiving module that receives user face image information and shooting environment condition information from a user terminal; a sensor module that collects at least one of an optical signal, a temperature signal, and an impedance signal from a skin surface; an on-device artificial intelligence module that is electrically connected to the image receiving module and the sensor module, calculates skin condition features using the face image information and the signal collected by the sensor module, and infers stimulation parameters corresponding to the calculated skin condition features; a stimulation module that is electrically connected to the on-device artificial intelligence module and outputs at least one of electrical stimulation, optical stimulation, high-frequency stimulation, and thermal stimulation according to the stimulation parameters; and a standardized metadata storage module that is electrically connected to the image receiving module, the sensor module, and the on-device artificial intelligence module, and generates standardized metadata by standardizing each of the face image information, the shooting environment condition information, the optical signal, the temperature signal, and the impedance signal according to an analysis reference value to compare the effect of improving the user's skin condition, and stores the generated standardized metadata.The system includes a cosmetic recommendation module that is electrically connected to the on-device artificial intelligence module and generates a list of cosmetics suitable for the user's skin based on the skin condition feature and provides it to the user terminal. The on-device artificial intelligence module sets individual skin condition reference features using the face capture image information and adjusts the stimulation parameters in a real-time closed-loop manner by comparing the skin condition feature after stimulation is applied to the skin surface by the stimulation module with the individual skin condition reference feature. It estimates the skin response pattern using a curve fitting algorithm, automatically optimizes the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters according to the estimated skin response pattern, predicts the possibility of overheating, overstimulation, and pain based on the impedance signal and the temperature signal to adjust the stimulation parameters, and automatically stops the stimulation output or lowers it to within a safety limit if contact failure or operational deviation is detected. The curve fitting algorithm adjusts the stimulation parameters by approximating the skin response to skin stimulation with a curve, and adapts to changes in individual skin sensitivity in a closed-loop manner by iteratively correcting the curve using the real-time skin response. The cosmetic recommendation module performs learning, receives information regarding the user's total income and fixed expenses from a user terminal to calculate disposable income, first filters cosmetics suitable for the user's skin according to the skin condition feature, and secondarily filters cosmetics among the first-filtered cosmetics that are less than or equal to a preset ratio of the disposable income to generate the cosmetic list, wherein the preset ratio is adjusted according to the user's gender and age group; Is This is achieved by a skin beauty device using artificial intelligence, characterized by the following.

[0011] In addition, the on-device artificial intelligence module can extract a skin analysis area from the face image information by using a face landmark detection model and a skin segmentation model.

[0012] In addition, the on-device artificial intelligence module can convert the color data of the face image information into the CIELAB color space, calculate brightness, redness, and yellowing values ​​to analyze the color and texture state of the skin, analyze wrinkles on the skin surface using a U-Net model, and detect blemishes and acne using a NanoDet model.

[0013] In addition, the on-device artificial intelligence module can estimate a skin response pattern using at least one of a Bayesian model, a multi-armed bandit algorithm, and a curve fitting algorithm, and can automatically optimize the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters according to the estimated skin response pattern.

[0014] The above other objective is, according to the present invention, an image receiving step of receiving user face image information and shooting environment condition information from a user terminal; a sensing step of collecting at least one of an optical signal, a temperature signal, and an impedance signal from a skin surface; a stimulation parameter inference step of calculating a skin condition feature using the face image information and the signal collected in the sensing step and inferring a stimulation parameter corresponding to the calculated skin condition feature; and a stimulation output step of outputting at least one of electrical stimulation, optical stimulation, and high-frequency stimulation according to the stimulation parameter.To compare the effect of improving the user's skin condition, the method includes a storage step of generating standardized metadata by standardizing each of the face capture image information, the capture environment condition information, the optical signal, the temperature signal, and the impedance signal according to an analysis reference value, and storing the generated standardized metadata; in the stimulation parameter inference step, individual skin condition reference features are set using the face capture image information, and the stimulation parameters are adjusted in a real-time closed-loop manner by comparing the image of the skin surface state after stimulation is applied to the skin surface by the stimulation output step with the individual skin condition reference features, wherein a skin response pattern is estimated using a curve fitting algorithm, and the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters are automatically optimized according to the estimated skin response pattern, and the stimulation parameters are adjusted by predicting the possibility of overheating, overstimulation, and pain based on the impedance signal and the temperature signal, and if contact failure or operation deviation is detected, the stimulation output is automatically stopped or lowered to within a safety limit, and the curve fitting algorithm adjusts the stimulation parameters by approximating the skin response to skin stimulation with a curve, and real-time skin This is achieved by a skin condition diagnosis method using artificial intelligence, characterized by performing closed-loop learning that adapts to individual changes in skin sensitivity by iteratively correcting the above curve using the response. Effects of the invention

[0015] According to the present invention, an on-device artificial intelligence module learns individual skin condition standard features and adjusts stimulation parameters in a real-time closed-loop manner so as to automatically set optimal stimulation according to the user's skin characteristics and session environment, thereby having the effect of applying skin stimulation optimized for the user.

[0016] In addition, according to the present invention, since only the area requiring skin analysis is accurately extracted from the entire face, the accuracy of skin analysis is significantly improved.

[0017] In addition, according to the present invention, since the user's skin surface condition is evaluated using quantified indicators, there is an effect of significantly improving the accuracy of tracking skin improvement and inferring stimulation parameters.

[0018] In addition, according to the present invention, since the skin response to stimulation is learned individually, the precision and efficiency of stimulation are improved.

[0019] Meanwhile, the effects of the present invention are not limited to those mentioned above, and various effects may be included within the scope obvious to a person skilled in the art from the contents described below. Brief explanation of the drawing

[0020] FIG. 1 illustrates the overall appearance of a skin beauty device using artificial intelligence according to an embodiment of the present invention, and FIG. 2 illustrates the configuration of a skin beauty device using artificial intelligence according to an embodiment of the present invention, and FIG. 3 illustrates the electrical connections between each component of a skin beauty device using artificial intelligence according to an embodiment of the present invention, and FIG. 4 is a flowchart of a skin condition diagnosis method using artificial intelligence according to an embodiment of the present invention. Specific details for implementing the invention

[0021] Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings.

[0022] In addition, when describing embodiments of the present invention, if it is determined that a detailed description of related known configurations or functions would hinder understanding of the embodiments of the present invention, such detailed description is omitted.

[0023] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the embodiments of the present invention. These terms are used merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by such terms.

[0024] Additionally, in describing embodiments of the present invention, each "part," "module," or "step" may be implemented through a processor and memory. The term "processor" should be broadly interpreted to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, and state machines, etc. In some environments, the term "processor" may refer to an application-specific integrated circuit (ASIC), a programmable logic device (PLD), a field programmable gate array (FPGA), etc. The term "processor" may also refer to a combination of processing devices, such as, for example, a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors combined with a DSP core, or any other combination of such configurations.

[0025] Additionally, memory should be broadly interpreted to include any electronic component capable of storing electronic information. Memory may also refer to various types of processor-readable media, such as random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), eraseable-programmable read-only memory (EPROM), electrically eraseable PROM (EEPROM), flash memory, magnetic or optical data storage devices, registers, etc. If a processor can read information from memory or write information to memory, memory is said to be in an electronic communication state with the processor, and each "part," "module," or "stage" may be implemented through a program or application based on the processor and memory in an electronic communication state.

[0026] In addition, the communication network may be a wired or wireless communication network and may utilize at least one of LAN (local area network), WLAN (wireless LAN), Wibro (wireless broadband), WiMAX (world interoperability for microwave access), HSDPA (high speed downlink packet access), and various other communication standards. Furthermore, the communication network may be a mobile communication network, and it may be a synchronous mobile communication network or an asynchronous mobile communication network. As an example of the asynchronous mobile communication network, a WCDMA (Wideband Code Division Multiple Access) network may be cited. In this case, although not shown in the drawings, the mobile communication network may include, for example, an RNC (Radio Network Controller). Meanwhile, although the WCDMA network was cited as an example, it may be a next-generation communication network such as a 3GLTE network, 4G network, or 5G network, as well as other low-power communication networks such as Wi-Fi, Bluetooth, and Zigbee, and wired / wireless IP networks.

[0027] Furthermore, in the present invention, Artificial Intelligence (AI) refers to a technology that imitates human learning ability, reasoning ability, and perceptual ability, and implements them on a computer, and may include concepts such as machine learning and symbolic logic. Machine Learning (ML) is an algorithmic technology that classifies or learns the characteristics of input information on its own. AI technology can analyze input information as a machine learning algorithm, learn from the results of that analysis, and make judgments or predictions based on the results of that learning. Additionally, technologies that mimic the functions of the human brain, such as cognition and judgment, by utilizing machine learning algorithms can also be understood as falling within the category of AI. For example, technological fields such as linguistic understanding, visual understanding, reasoning / prediction, knowledge representation, and motion control may be included.

[0028] Artificial intelligence learning models or neural network models can be designed to implement the structure of the human brain on a computer and may include multiple network nodes that simulate neurons of a human neural network and have weights. These multiple network nodes can have interconnected relationships by simulating the synaptic activity of neurons, where neurons exchange signals through synapses. In an artificial intelligence learning model, multiple network nodes can be located in layers of different depths and exchange information according to convolutional connections. An artificial intelligence learning model may be, for example, an Artificial Neural Network (ANN) or a Convolutional Neural Network (CNN). An artificial intelligence learning model can be machine learned according to methods such as supervised learning, unsupervised learning, and reinforcement learning. Machine learning algorithms that can be used to perform machine learning include Decision Trees, Bayesian Networks, Support Vector Machines, Perceptrons, and Clustering.

[0030] From now on, with reference to the attached drawings, a skin beauty device (100) using artificial intelligence according to an embodiment of the present invention will be described in detail.

[0031] FIG. 1 illustrates the overall appearance of a skin beauty device using artificial intelligence according to an embodiment of the present invention, FIG. 2 illustrates the configuration of a skin beauty device using artificial intelligence according to an embodiment of the present invention, and FIG. 3 illustrates the electrical connections between each component of a skin beauty device using artificial intelligence according to an embodiment of the present invention.

[0032] As illustrated in FIGS. 1 to 3, a skin beauty device (100) using artificial intelligence according to an embodiment of the present invention may include an image receiving module (110), a sensor module (120), an on-device artificial intelligence module (130), a stimulation module (140), and / or a standardized metadata storage module (150).

[0033] The image receiving module (110) receives information on the user's face image and shooting environment conditions from the user terminal, is connected to communicate with the user terminal, and is electrically connected to the on-device artificial intelligence module (130) described later.

[0034] Here, the user terminal includes smartphones, tablets, action cams, etc., but is not limited thereto, and any device capable of taking photos or videos and transmitting captured image information to the image receiving module (110) is acceptable.

[0035] The face capture image information may include image color data, and the shooting environment condition information may include data regarding shooting environment conditions reflected in capturing the user's face, such as ISO sensitivity, shutter speed, white balance, lighting intensity, shooting posture, and shooting focal length.

[0036] Meanwhile, the user's face image information and shooting environment condition information captured by the user terminal can be used by the on-device artificial intelligence module (130) described later to calculate the user's skin condition features, which will be described later.

[0037] The sensor module (120) collects at least one of an optical signal, a temperature signal, and an impedance signal from the skin surface and is electrically connected to the on-device artificial intelligence module (130) described later.

[0038] The sensor module (120) may include, more specifically, a light sensor, a temperature sensor and / or an impedance sensor.

[0039] The optical sensor can collect optical signals by measuring the color, intensity, and wavelength distribution of light reflected from the skin surface and light scattered within the skin, respectively, through an LED light that irradiates light onto the skin surface.

[0040] The temperature sensor is provided as a sensor capable of measuring the temperature of the skin surface, such as a thermistor or an infrared sensor, and can collect a temperature signal by detecting the skin surface temperature.

[0041] An impedance sensor can collect an impedance signal by passing an electric current through the skin surface and measuring electrical resistance. The impedance signal can be used to analyze the skin's moisture content, keratinization condition, and the like.

[0042] Meanwhile, the optical signal collected by the light sensor, the temperature signal collected by the temperature sensor, and the impedance signal collected by the impedance sensor can be used by the on-device artificial intelligence module (130) described later to calculate skin condition features, which will be described later.

[0043] The on-device artificial intelligence module (130) calculates skin condition features using face image information and signals collected by the sensor module (120), and infers stimulation parameters corresponding to the calculated skin condition features, and is electrically connected to the image receiving module (110), the sensor module (120), the stimulation module (140) described later, and the standardized metadata storage module (150).

[0044] More specifically, the on-device artificial intelligence module (130) operates independently from an external cloud server and, using lightweight artificial intelligence models such as TFLite and ONNX, can recognize the user's face from the face capture image information received by the image receiving module (110), and analyze the user's skin condition by analyzing the shooting environment condition information and the optical signal, temperature signal, and impedance signal collected by the sensor module (120), thereby producing skin condition features.

[0045] The skin condition feature is an indicator representing the user's skin condition, and may include L*, a*, and b* values ​​of a face image taken by the user through a user terminal after skin stimulation is applied by the stimulation module (140) described later, converted into a CIELAB color space, the color, intensity, and wavelength distribution of light reflected from the skin surface and light scattered within the skin, respectively, the skin surface temperature value, and the electrical resistance value of the skin.

[0046] Stimulation parameters refer to detailed variables for controlling each of electrical muscle stimulation (EMS), photo-stimulation, high-frequency stimulation, and thermal stimulation. According to the stimulation parameters, the frequency, waveform, duty cycle, and current of electrical stimulation, the wavelength, light intensity, and pulse of photo-stimulation, the output and duty cycle of high-frequency stimulation, and the cooling and heating times of thermal stimulation can be adjusted, respectively. Here, the duty cycle refers to the ratio of the operating time during one cycle.

[0047] Additionally, the on-device artificial intelligence module (130) can set individual skin condition reference features using face image information, and by comparing the skin condition features after stimulation is applied to the skin surface by the stimulation module (140) with the individual skin condition reference features, the stimulation parameters can be adjusted in a real-time closed-loop manner.

[0048] The individual skin condition standard feature is intended to compare changes in skin condition after skin stimulation is applied by the stimulation module (140) described later, and may include information on face image information and shooting environment conditions taken by the user for the first time or at the beginning of the session through the user terminal, as well as information on optical signals, temperature signals, and impedance signals collected through the sensor module (120). Here, a session refers to a unit of usage when the same user repeatedly uses the skin beauty device.

[0049] In addition, the on-device artificial intelligence module (130) can extract a skin analysis area from face image information using a face landmark detection model and a skin segmentation model.

[0050] The face landmark detection model can detect the user's face region from face capture image information and extract face geometric information including coordinate information of major feature points within the face, such as eyes, nose, mouth, jawline, and face contour. The face landmark detection model can be implemented as a MediaPipe Face Landmarker model, but is not limited thereto.

[0051] The skin segmentation model can use facial geometric information extracted by the face landmark detection model to subdivide the face region into pixel units and generate a binary mask representing the skin region excluding non-skin regions such as eyes, mouth, and hair. The skin segmentation model can be implemented as a face parsing model, but is not limited thereto.

[0052] Additionally, the on-device artificial intelligence module (130) can convert color data of face image information into the CIELAB color space, calculate brightness, redness, and yellowing values ​​to analyze the color and texture state of the skin, analyze wrinkles on the skin surface using a U-Net model, and detect blemishes and acne using a NanoDet model.

[0053] In the CIELAB color space, color coordinates are distinguished as L*, a*, and b*; L* represents brightness, a* represents the degree of bias toward red or green, and b* represents the degree of bias toward yellow or blue.

[0054] In other words, when facial image information is converted to the CIELAB color space, the brightness and gloss of the skin can be analyzed based on the L* value, the degree of erythema (redness) of the skin can be analyzed based on the a* value, and the degree of yellowness of the skin can be analyzed based on the b* value.

[0055] The U-Net model is a neural network with an encoder-decoder symmetric structure that can analyze skin texture by segmenting fine wrinkle patterns from captured facial image information at the pixel level and analyzing the location, length, depth, and direction of wrinkles and pores. In this case, the captured facial image input to the U-Net model may have undergone wrinkle bandpass processing by converting it to grayscale and filtering it using a Difference of Gaussian (DoG) or Frangi filter.

[0056] The NanoDet model is a lightweight object detection network capable of detecting lesion areas, such as acne, blemishes, and melasma, from captured facial images at the pixel level. The NanoDet model features a one-stage detector structure based on a Feature Pyramid Network (FPN) and can operate in real-time in an on-device environment due to its low computational load. Information regarding the location, size, and number of lesions detected by the NanoDet model can be utilized as skin condition features and used for adjusting stimulation parameters and analyzing skin improvement effects.

[0057] Additionally, the on-device artificial intelligence module (130) can estimate a skin response pattern using at least one of a Bayesian model, a multi-armed bandit algorithm, and a curve fitting algorithm, and can automatically optimize the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters according to the estimated skin response pattern.

[0058] Bayesian models probabilistically model the relationship between skin irritation and skin response to estimate the distribution of the effects of each stimulus on the skin condition; when skin condition features and irritation parameters are input, they can probabilistically predict skin response quantities such as skin tone improvement rate, moisture recovery rate, and erythema change rate.

[0059] The Multi-Armed Bandit algorithm is a sequential decision-making algorithm that solves the probabilistic reward problem and can infer optimal stimulation parameters by evaluating the probability of skin response for each combination of stimulation parameters, such as electrical stimulation, photo-stimulation, high-frequency stimulation, and thermal stimulation.

[0060] The curve fitting algorithm is designed to correct stimulation parameters by estimating a continuous functional relationship between skin stimulation and skin response. It automatically adjusts stimulation parameters by approximating skin responses—such as changes in erythema, moisture, and temperature increase resulting from skin stimulation (e.g., current, frequency, and light intensity)—as curves. Furthermore, by iteratively correcting the curves using real-time skin responses, it can perform closed-loop learning that adapts to individual changes in skin sensitivity.

[0061] That is, the on-device artificial intelligence module (130) can detect skin brightness, erythema, yellowing, wrinkle depth, wrinkle length, wrinkle direction, blemishes, acne, pore location, pore area, etc. by using a face landmark detection model, skin segmentation model, CIELAB color space conversion, U-Net model, NanoDet model, Bayesian model, multi-armed bandit algorithm and curve fitting algorithm as described above, and can analyze skin color, skin texture and skin lesions, and can infer stimulation parameters suitable for the user's skin according to the prediction of skin response based on the calculated skin condition features and stimulation parameters, and can infer user-customized stimulation parameters by automatically repeating and learning stimulation parameters according to real-time skin response.

[0062] For example, if the on-device artificial intelligence module (130) determines that the user's skin blood flow is reduced based on the calculated user's skin condition features, it can adjust the stimulation parameters to increase the frequency and duty of the electrical stimulation, and the stimulation module (140) described later can improve the skin blood flow by stimulating the skin according to the adjusted stimulation parameters, and through the real-time skin response according to the adjusted stimulation parameters, it can reduce the intensity of the light stimulation to prevent the skin from being excessively stimulated.

[0063] Additionally, the on-device artificial intelligence module (130) can predict the possibility of overheating, overstimulation, or pain based on impedance signals, temperature signals, and usage history, and can adjust stimulation parameters, and if abnormal signs such as poor contact or malfunction are detected, it can automatically stop the stimulation output or immediately lower it to within safe limits.

[0064] For example, if the amount of temperature change on the skin surface during use of the skin beauty device exceeds a preset reference value, the on-device artificial intelligence module (130) can lower or block the stimulation parameter of the heat stimulation to prevent excessive stimulation from being applied to the skin.

[0065] According to this, since excessive stimulation to the skin can be prevented in advance, the safety of using skin beauty devices can be improved compared to rule-based stimulation parameter adjustment methods.

[0066] Additionally, the on-device artificial intelligence module (130) can generate stimulation parameters that reflect the transmitted preset when an administrator sets a preset and transmits it to a skin beauty device, generate a skin reaction result based on the stimulation parameters that reflect the transmitted preset as a usage log and transmit it back to an external terminal or clinic system, and optimize the stimulation parameters for the next session based on the generated usage log.

[0067] That is, the stimulation parameter with the preset can be applied to the skin by the stimulation module (140), and the resulting skin condition feature can be compared and analyzed with the individual skin condition standard feature and used to infer the stimulation parameter in the next session.

[0068] In addition, the on-device artificial intelligence module (130) can graph the stimulation parameters and changes in skin condition after stimulation for each user, stimulation area, and stimulation time, and structure them in the form of a time series of time flows. According to this, by directly learning the Dose-Response curve representing the stimulation-response causal relationship through the accumulation of session logs, the learning efficiency is improved, personalized stimulation parameter inference is possible, long-term skin improvement trends can be analyzed by tracking changes in skin condition between sessions, and stimulation parameters for the next session can be optimized in real time.

[0069] The stimulation module (140) outputs at least one of electrical stimulation, light stimulation, high-frequency stimulation, and thermal stimulation according to stimulation parameters and is electrically connected to the on-device artificial intelligence module (130).

[0070] The stimulation module (140) may more specifically include an EMS stimulation unit, an LED stimulation unit, an RF stimulation unit and / or a thermal control unit, and each of the EMS stimulation unit, LED stimulation unit, RF stimulation unit and thermal control unit may apply electrical stimulation, photo stimulation, high-frequency stimulation, and thermal stimulation to the skin surface according to stimulation parameters inferred by the on-device artificial intelligence module (130).

[0071] Specifically, the EMS stimulation unit can induce lifting and elasticity improvement by applying electrical stimulation to the epidermal and dermal muscle layers, the LED stimulation unit can induce cell metabolism promotion, erythema relief, and pigmentation improvement by irradiating the skin surface with light of wavelengths in the visible or near-infrared region, the RF stimulation unit can induce collagen contraction and regeneration by heating the dermal layer through high-frequency power, and the heat control unit can improve blood flow by cooling the skin after stimulation through a Peltier element or heating the skin through a heating element.

[0072] The standardized metadata storage module (150) generates standardized metadata by standardizing each of the face shooting image information, shooting environment condition information, optical signal, temperature signal, and impedance signal according to an analysis reference value to compare the effect of improving the user's skin condition, and stores the generated standardized metadata, and is electrically connected to the image receiving module (110), sensor module (120), and on-device artificial intelligence module (130).

[0073] The standardized metadata storage module (150) can, more specifically, calculate the average values ​​of L*, a*, and b* values ​​of the face image information converted into the CIELAB color space by the on-device artificial intelligence module (130), as well as the average values ​​of the optical signal, temperature signal, and impedance signal, and normalize shooting environment condition information such as ISO sensitivity, shutter speed, white balance, lighting intensity, shooting posture, and shooting focal length according to the same standard to generate a standardized data set that enables comparison of skin conditions between sessions. For example, if the ISO sensitivity of the face image taken at the start of the first session is 200 and the ISO sensitivity of the face image taken at the start of the second session is 400, the standardized metadata storage module (150) can correct the ISO sensitivity of the second face image to be 200 so that the face images between sessions can be compared under the same conditions.

[0074] The generated standardized metadata is structured and stored by session, area, and time, and can be used by the aforementioned on-device artificial intelligence module (130) to quantitatively evaluate the degree of skin improvement by comparing the difference with the skin condition features obtained in subsequent sessions.

[0075] In addition, the generated standardized metadata can be used as reference data necessary for the aforementioned on-device artificial intelligence module (130) to analyze long-term improvement trends, such as changes in skin tone, degree of wrinkle reduction, and moisture recovery, by accumulating and managing long-term data of the same user, and to correct stimulation parameters in a closed-loop manner.

[0076] That is, the standardized metadata storage module (150) can perform the role of improving the reliability and repeatability of AI learning by standardizing various environmental variables and biosignals according to analysis reference values ​​to enable quantitative comparison between sessions.

[0077] According to a skin beauty device (100) using artificial intelligence according to an embodiment of the present invention, comprising an image receiving module (110), a sensor module (120), an on-device artificial intelligence module (130), a stimulation module (140), and / or a standardized metadata storage module (150) as described above, the on-device artificial intelligence module (130) learns individual skin condition standard features and adjusts stimulation parameters in a real-time closed-loop manner so as to automatically set optimal stimulation according to the user's skin characteristics and session environment, thereby having the effect of applying skin stimulation optimized for the user. In addition, since only the area requiring skin analysis is accurately extracted from the entire face, the accuracy of skin analysis is greatly improved. In addition, since the user's skin surface condition is evaluated as a quantified indicator, the accuracy of tracking skin improvement and inferring stimulation parameters is greatly improved. Furthermore, since the skin response to stimulation is learned individually, the precision and efficiency of stimulation are improved.

[0079] Meanwhile, in order to improve the user's skin condition, there are cases where not only must physical stimulation be applied to the skin through a skin beauty device, but cosmetics suitable for the user's skin condition must also be used.

[0080] Accordingly, a skin beauty device using artificial intelligence according to one embodiment of the present invention may further include a cosmetic recommendation module (160).

[0081] The cosmetic recommendation module (160) recommends cosmetics suitable for the skin based on skin condition features calculated by the on-device artificial intelligence module (130), is linked with the user terminal, and is electrically connected to the on-device artificial intelligence module.

[0082] More specifically, the cosmetic recommendation module (160) is linked to a user terminal, receives information regarding the user's total income and fixed expenses from the user terminal, calculates disposable income by subtracting fixed expenses from the user's total income, is linked to an online cosmetic shopping mall server through the user terminal, first filters a list of cosmetics suitable for the skin according to the user's skin condition feature, second filters a list of cosmetics among the first filtered cosmetic list that corresponds to a predetermined percentage amount or less of the calculated disposable income, and can provide the second filtered cosmetic list to the user terminal.

[0083] Here, fixed expenses include, but are not limited to, monthly rent, insurance premiums, loan principal and interest, installment payments, taxes, maintenance fees, communication charges, and regular donations.

[0084] In addition, a predetermined percentage applied to the calculated disposable income may be set differently depending on gender and age group. For example, it may be 5% for men in their 20s, 10% for men in their 30s, 7% for men in their 40s, 10% for women in their teens and 20s, 20% for women in their 30s, 15% for women in their 40s, and 10% for women in their 50s or older, but is not limited thereto. This is because, generally, women have a higher interest in skin care than men, and younger age groups may have a higher interest in skin care compared to other age groups.

[0085] That is, the cosmetic recommendation module (160) can, for example, when a user of the skin beauty device (100) using artificial intelligence according to one embodiment of the present invention is a woman in her 30s with a total income of 4 million won and fixed expenses of 2.2 million won, and the L* value of a face image converted to the CIELAB color space by the on-device artificial intelligence module (130) is measured as 35 and recommends whitening cosmetics for skin brightness improvement, it can first filter the list of whitening cosmetics among the list of cosmetics sold in an online cosmetic shopping mall, and secondly filter the list of whitening cosmetics priced at 360,000 won or less, which corresponds to a 20% ratio of the disposable income of 1.8 million won (total income of 4 million won minus fixed expenses of 2.2 million won), and provide the secondly filtered list of whitening cosmetics to the user terminal.

[0086] According to this cosmetic recommendation module (160), there is an effect of automatically recommending the optimal cosmetic for improving skin condition based on individual income level and individual skin condition features that reflect fixed expenses.

[0088] From now on, with reference to the attached drawings, a skin diagnosis method (S100) using artificial intelligence according to an embodiment of the present invention will be described in detail.

[0089] FIG. 4 is a flowchart of a skin condition diagnosis method using artificial intelligence according to an embodiment of the present invention.

[0090] As illustrated in FIG. 4, a skin diagnosis method (S100) using artificial intelligence according to an embodiment of the present invention may include an image receiving step (S110), a sensing step (S120), a stimulation parameter inference step (S130), a stimulation output step (S140) and / or a storage step (S150).

[0091] The image reception step (S110) is a step of receiving user face image information and shooting environment condition information from a user terminal, and is performed by the aforementioned image reception module (110).

[0092] The sensing step (S120) is a step of collecting at least one of an optical signal, a temperature signal, and an impedance signal from the skin surface, and is performed by the aforementioned sensor module (120).

[0093] The stimulation parameter inference step (S130) is a step of calculating skin condition features using face image information and signals collected in the sensing step (S120), and inferring stimulation parameters corresponding to the calculated skin condition features, and is performed by the aforementioned on-device artificial intelligence module (130).

[0094] In addition, in the stimulation parameter inference step (S130), individual skin condition reference features are set using face image information, and by comparing the image of the skin surface condition after stimulation is applied to the skin surface by the stimulation output step (S140) with the individual skin condition reference features, the stimulation parameters can be adjusted in a real-time closed-loop manner.

[0095] The stimulation output step (S140) is a step of outputting at least one of electrical stimulation, optical stimulation, and high-frequency stimulation according to stimulation parameters, and is carried out by the aforementioned stimulation module (140).

[0096] The storage step (S150) is a step of generating standardized metadata by standardizing each of the face shooting image information, shooting environment condition information, optical signal, temperature signal, and impedance signal according to an analysis reference value to compare the effect of improving the user's skin condition, and storing the generated standardized metadata, and is carried out by the aforementioned standardized metadata storage module (150).

[0097] Meanwhile, the image receiving module (110) for performing the image receiving step (S110), the sensor module (120) for performing the sensing step (S120), the on-device artificial intelligence module (130) for performing the stimulus parameter inference step (S130), the stimulus module (140) for performing the stimulus output step (S140), and the standardized metadata storage module (150) for performing the storage step (S150) are each as described above, so a redundant description is omitted.

[0098] According to a skin diagnosis method (S100) using artificial intelligence according to an embodiment of the present invention, comprising an image reception step (S110), a sensing step (S120), a stimulation parameter inference step (S130), a stimulation output step (S140), and / or a storage step (S150) as described above, an on-device artificial intelligence module learns individual skin condition reference features and adjusts stimulation parameters in a real-time closed-loop manner so as to automatically set optimal stimulation according to the user's skin characteristics and session environment, thereby having the effect of applying skin stimulation optimized for the user. In addition, since only the area requiring skin analysis is accurately extracted from the entire face, the accuracy of skin analysis is significantly improved. Furthermore, since the user's skin surface condition is evaluated using quantified indicators, the accuracy of tracking skin improvement and inferring stimulation parameters is significantly improved. In addition, since the skin response to stimulation is learned individually, the precision and efficiency of stimulation are improved.

[0100] Although all components constituting the embodiments of the present invention have been described above as being combined or operating in combination, the present invention is not necessarily limited to such embodiments. That is, within the scope of the purpose of the present invention, all components may be selectively combined in one or more ways to operate.

[0101] Furthermore, terms such as "include," "compose," or "have" described above, unless specifically stated otherwise, mean that the relevant component may be inherent; therefore, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the present invention.

[0102] Furthermore, the above description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention.

[0103] Accordingly, the embodiments disclosed in this invention are intended to illustrate, not limit, the technical concept of the invention, and the scope of the technical concept of the invention is not limited by these embodiments. The scope of protection of this invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of this invention. Explanation of the symbols

[0104] 100: Skin beauty device using artificial intelligence according to an embodiment of the present invention 110: Image receiving module 120 : Sensor module 130: On-device AI Module 140: Stimulation Module 150: Standardized metadata storage module 160: Cosmetics Recommendation Module S100: A method for diagnosing skin condition using artificial intelligence according to an embodiment of the present invention S110: Image reception step S120: Sensing stage S130: Stimulus parameter inference step S140: Stimulation output stage S150: Storage step

Claims

Claim 1 An image receiving module that receives user face image information and shooting environment condition information from a user terminal; a sensor module that collects at least one of an optical signal, a temperature signal, and an impedance signal from a skin surface; an on-device artificial intelligence module that is electrically connected to the image receiving module and the sensor module, calculates skin condition features using the face image information and the signal collected by the sensor module, and infers stimulation parameters corresponding to the calculated skin condition features; a stimulation module that is electrically connected to the on-device artificial intelligence module and outputs at least one of electrical stimulation, photo stimulation, high-frequency stimulation, and thermal stimulation according to the stimulation parameters; and a standardized metadata storage module that is electrically connected to the image receiving module, the sensor module, and the on-device artificial intelligence module, and generates standardized metadata by standardizing each of the face image information, the shooting environment condition information, the optical signal, the temperature signal, and the impedance signal according to an analysis reference value to compare the effect of improving the user's skin condition, and stores the generated standardized metadata.The system includes a cosmetic recommendation module that is electrically connected to the on-device artificial intelligence module and generates a list of cosmetics suitable for the user's skin according to the skin condition feature and provides it to the user terminal. The on-device artificial intelligence module sets individual skin condition reference features using the face capture image information and adjusts the stimulation parameters in a real-time closed-loop manner by comparing the skin condition feature after stimulation is applied to the skin surface by the stimulation module with the individual skin condition reference feature. It estimates the skin response pattern using a curve fitting algorithm and automatically optimizes the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters according to the estimated skin response pattern. It also predicts the possibility of overheating, overstimulation, and pain based on the impedance signal and the temperature signal to adjust the stimulation parameters, and automatically stops the stimulation output or lowers it to within a safety limit if contact failure or operational deviation is detected. The curve fitting algorithm adjusts the stimulation parameters by approximating the skin response to skin stimulation with a curve and adapts to changes in individual skin sensitivity through a closed-loop method by iteratively correcting the curve using the real-time skin response. The cosmetic recommendation module performs learning, receives information regarding the user's total income and fixed expenses from a user terminal to calculate disposable income, first filters cosmetics suitable for the user's skin according to the skin condition feature, and secondarily filters cosmetics among the first-filtered cosmetics that are less than or equal to a preset ratio of the disposable income to generate the cosmetic list, wherein the preset ratio is adjusted according to the user's gender and age group; Is A skin beauty device using artificial intelligence characterized by the following. Claim 2 A skin beauty device using artificial intelligence according to claim 1, wherein the on-device artificial intelligence module extracts a skin analysis area from the face capture image information using a face landmark detection model and a skin segmentation model. Claim 3 A skin beauty device using artificial intelligence according to claim 1, wherein the on-device artificial intelligence module converts color data of the face capture image information into the CIELAB color space, calculates brightness, redness, and yellowing values ​​to analyze the color and texture state of the skin, analyzes wrinkles on the skin surface using a U-Net model, and detects blemishes and acne using a NanoDet model. Claim 4 delete Claim 5 An image receiving step for receiving user face image information and shooting environment condition information from a user terminal; a sensing step for collecting at least one of an optical signal, a temperature signal, and an impedance signal from a skin surface; a stimulation parameter inference step for calculating a skin condition feature using the face image information and the signal collected in the sensing step, and inferring a stimulation parameter corresponding to the calculated skin condition feature; and a stimulation output step for outputting at least one of electrical stimulation, photo stimulation, and high-frequency stimulation according to the stimulation parameter.To compare the effect of improving the user's skin condition, the method includes a storage step of generating standardized metadata by standardizing each of the face image information, the shooting environment condition information, the optical signal, the temperature signal, and the impedance signal according to an analysis reference value, and storing the generated standardized metadata; in the stimulation parameter inference step, individual skin condition reference features are set using the face image information, and the stimulation parameters are adjusted in a real-time closed-loop manner by comparing an image of the skin surface state after stimulation is applied to the skin surface by the stimulation output step with the individual skin condition reference features, wherein a skin response pattern is estimated using a curve fitting algorithm, and the stimulation intensity, stimulation time, and stimulation waveform of the stimulation parameters are automatically optimized according to the estimated skin response pattern; the stimulation parameters are adjusted by predicting the possibility of overheating, overstimulation, and pain based on the impedance signal and the temperature signal, and if contact failure or operational deviation is detected, the stimulation output is automatically stopped or lowered to within a safety limit; the curve fitting algorithm adjusts the stimulation parameters by approximating the skin response to skin stimulation with a curve, and real-time A method for diagnosing skin condition using artificial intelligence, characterized by performing closed-loop learning that adapts to individual changes in skin sensitivity by iteratively correcting the curve using skin response.

Images